OK, I catch your drift.
What you are asking for is a bit tricky. I'd implement it with Text Elements and conditionals, but it only gets messier from there...
You'd HAVE to use text elements, as there is no way to tell what the menu items will be (unless it was a custom, limited choice menu, which you could simulate what you want now) - remember, skins are supposed to be generic and there are hundreds of games and apps one might have on their system.
The next problem is navigating submenus, and handling situations where there are more items than visible "positions" displayed.
Perhaps with integration with ActionScripting, it might be possible. I'll give that some thought.
As for the here and now, if you have, say, a limited 3 item menu, and wanted to use images, you could simulate it with a 1 row menu and use the next/prev conditionals to simulate the choice currently available and "trick" the system into displaying the items you wanted to. Furthermore, the 1 row menu can be placed offscreen, so as to avoid it's part in the display.